Monetary policy and household consumption
Empirical evaluations of monetary policy have traditionally focused on the responses of macroeconomic aggregates. Instead, this column uses detailed administrative data from Norway to uncover substantial heterogeneity in the effects of monetary policy at the household level. The authors find that not only low-liquidity households but also high-liquidity ones also show strong responses. Interest rate changes faced by borrowers and savers feed into consumption, and indirect effects of monetary policy are sizable, but occur with a delay. While the results confirm several predictions of recent heterogeneous-agent New Keynesian models, they also provide new challenges.
There is a wide consensus that interest rate changes by central banks have an impact on aggregate consumption. However, several aspects of this link cannot be assessed based on macroeconomic aggregates alone and have therefore remained in the dark for a long time. Through which channels are households affected by policy rate changes? Is there significant heterogeneity in their reactions? What is the role of household portfolios and borrowing constraints? While these questions have received considerable attention in recent theoretical contributions (e.g. Kaplan et al. 2018,Bilbiie 2019, Luetticke 2019, Auclert et al. 2020, among many others), empirical evidence has proved difficult to obtain. Studies based on household surveys have yielded a number of valuable insights. For example, Cloyne et al. (2019) find that the balance sheets of households, approximated by their housing tenure status, affect the consumption response to monetary policy shocks, and Wong (2019) emphasises the importance of mortgage refinancing.
In recent research (Holm et al. 2020), we use administrative tax data covering all individuals registered in Norway between 1996 and 2015 to shed light on the effects of monetary policy at the household level. Since Norway levies both income and wealth taxes on its inhabitants, the data contain detailed records of income flows and portfolio positions, which allows us to impute the consumption expenditures of households. Equipped with these unique data and new estimates of monetary policy shocks for Norway, we are able to estimate the dynamic responses to monetary policy at a high level of disaggregation.
Liquid assets and heterogeneity in the consumption response
A number of recent HANK models predict that households differ markedly in their response to monetary policy and that these differences are related to holdings of liquid assets like bank deposits, which can be withdrawn quickly and at little or no cost. Intuitively, households with few liquid assets do not have a large enough buffer to smooth consumption when interest rate changes let their income fluctuate. If these households face borrowing costs or constraints, then their consumption is affected particularly strongly.
Figure 1 Impulse responses of consumption along the liquid asset distribution
Notes: Consumption changes normalized by lagged net income.
Source: Holm et al. (2020)
Estimates based on the household-level data confirm that the consumption response differs with the liquidity position, as can be seen from Figure 1. The figure illustrates the impulse responses of consumption to a monetary tightening for each decile of households ordered by their liquid asset holdings. Shown are time-averaged versions of the full responses. On impact and in the year following the shock, consumption declines below the 70th percentile of the distribution. The decline after four to five years is stronger at the bottom than around the median. But, strikingly, households with large liquid asset holdings also show a strong reaction. Moreover, they initially increase their consumption before eventually letting it decline.
Monetary policy affects consumption through income changes
The consumption response can be separated into the responses of income and saving, since all disposable household income in a given time period must be either consumed or saved. This is done in the two panels at the top of Figure 2. The two panels at the bottom separate the disposable income response into financial and nonfinancial income.
Figure 2 Impulse responses of income and saving along the liquid asset distribution
Notes: Changes in income and saving normalized by lagged net income.
Source: Holm et al. (2020)
Overall, the income response is large compared to the saving response, suggesting that the consumption adjustments that households make are closely linked to changes in their income. At the bottom of the distribution, households do not cut into their savings or increase their borrowing when disposable income starts to decline about two years after the shock. While this finding could be a result of liquidity shortage, as suggested by recent HANK models, the strong response at the top of the distribution is difficult to reconcile with explanations of monetary policy transmission that focus on liquidity constraints. According to these explanations, house- holds with large liquid asset holdings should smooth temporary income fluctuations by adjusting their saving and intertemporally substitute consumption in response to interest rate changes. In contrast, the estimates suggest that even high-liquidity households have sizable marginal propensities to consume out of income changes caused by monetary policy shocks, since both consumption and saving rise when interest rates and disposable income increase in years 0 to 1.
The initial increase in disposable income, and hence consumption, at the top of the distribution, and the decline towards the bottom are driven by the respective net financial income responses. Households with large liquidity holdings tend to be net savers whose interest income increases when interest rates rise. Households at the bottom tend to be net debtors with large mortgage debt. Their disposable income initially declines as the interest expenses associated with their mortgages increase. In our paper, we show that these differences are even more pronounced when households are reordered according to their net interest rate exposure. Nonfinancial income – the sum of labour earnings and net transfers – responds sluggishly but ultimately falls across the entire distribution.
Indirect effects are sizable but delayed
The strong nonfinancial income response suggests that general equilibrium effects play an important role in transmitting changes in the policy target rate to household consumption. Kaplan et al. (2018) show that such indirect effects of monetary policy make up a large fraction of the total consumption response in their HANK framework. Figure 3 provides a de- composition of the aggregate effect into direct and indirect effects based on estimates from the administrative data. The blue line gives the total response across all households, and the red line portrays the response when general equilibrium income changes are held constant, that is, when monetary policy affects households only in a direct way.
Figure 3 Direct and indirect effects of monetary policy
Notes: Impulse responses to a 1 percentage point contractionary shock. 68 percent confidence bands based on Driscoll- Kraay strd. errors shown.
Source: Holm et al. (2020)
In the first few years, almost all of the consumption response is due to direct effects. The gap between both curves subsequently widens, implying that indirect effects play an increasingly important role. Five years after the shock, more than half of the total response is due to indirect effects.
Taken together, the administrative data yield a number of new insights that may guide policy makers and future HANK models alike. There is large heterogeneity in the response of household consumption expenditures to monetary policy changes. Even the sign of the initial response differs for net creditors and net debtors. Households at the top of the liquid asset distribution, which should not be affected by borrowing constraints, show sizable marginal propensities to consume out of the income changes caused by policy rate adjustments. Indirect effects are important, but it takes time for general equilibrium effects to take over, suggesting that other forces have to respond first to initiate them.
Auclert, A, M Rognlie, and L Straub (2020), “Micro jumps, macro humps: monetary policy and business cycles in an estimated HANK model”, NBER Working Paper 26647.
Bilbiie, F O (2019), “The New Keynesian cross”, Journal of Monetary Economics, forthcoming.
Cloyne, J, C Ferreira, and P Surico (2019), “ Monetary policy when households have debt: New evidence on the transmission mechanism”, The Review of Economic Studies 87 (1): 102–129.
Holm, M B, P Paul, and A Tischbirek (2020), “The Transmission of Monetary Policy under the Microscope”, Federal Reserve Bank of San Francisco Working Paper 2020-03.
Kaplan, G, B Moll, and G L Violante (2018), “Monetary Policy According to HANK”, American Economic Review 108 (3): 697–743.
Luetticke, R (2019), “Transmission of monetary policy with heterogeneity in household portfolios”, Unpublished manuscript.
Wong, A (2019), “Refinancing and the Transmission of Monetary Policy to Consumption”, Unpublished manuscript.